Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Re: Reporting Gamma Regression


From   Roger Newson <roger.newson@kcl.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Re: Reporting Gamma Regression
Date   Mon, 09 Aug 2004 17:15:29 +0100

I would exponentiate the coefficients using -eform-, and describe them as arithmetic means and/or arithmetic mean ratios. These ratios may be ratios between 2 group arithmetic means, or ratios caused by a unit increment in a continuous X-variable. They are similar to (but not identical to) the geometric means and geometric mean ratios measured by the eform() option of -regress- with logged data. (My Stata tip on geometric mean ratios can be downloaded in Stata by typing

findit gmratio

and downloading the tip in .pdf format.) Arithmetic mean ratios have the advantage over geometric mean ratios that they still exist for variables with a nonzero probability of being zero.

I hope this helps.

Roger

At 16:44 09/08/2004, Joseph Coveney wrote:

Stephen Soldz wrote:

> A non-Stata question, though I'm using Stata for the anlysis. I'm analysing
> cost data and have used a GLM with gamma family and log link. The question I
> have is regards reporyting results. Should I exponentiate the coefficients
> (eform)? If so, any suggestions on how to describe this? Does anyone know of
> published examples I could refer to?

Did you ever get an answer?

The examples that crop up on the Web tend to either leave the coefficients
unexponentiated ( http://harrisschool.uchicago.edu/pdf/wp_03_13.pdf this is
Willard Manning, Anirban Basu & John Mullahy) or seem to use predictions (mu)
and talk about changes in them
( www.journals.uchicago.edu/CID/journal/issues/v33n12/001639/001639.text.html
and www.dianthus.co.uk/resources/sampleReport.pdf , and maybe
www.academyhealth.org/arm/abstracts/medicare.pdf ).

Log-gamma models are discussed in J. Hardin & J. Hilbe, _Generalized Linear
Models and Extensions_ (College Station, Texas: Stata Corporation, 2001),
pp. 69-71. They, too, didn't exponentiate the coefficients.

The one example of exponentiated coefficients that turned up (not exhaustive)
( www.mrc-bsu.cam.ac.uk/BSUsite/Publications/Preprints/glm_p6.pdf ) didn't give
them a very snazzy name, analogous to risk ratios or incidence rate ratios.

Joseph Coveney


*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/

--
Roger Newson
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
United Kingdom

Tel: 020 7848 6648 International +44 20 7848 6648
Fax: 020 7848 6620 International +44 20 7848 6620
  or 020 7848 6605 International +44 20 7848 6605
Email: roger.newson@kcl.ac.uk
Website: http://www.kcl-phs.org.uk/rogernewson

Opinions expressed are those of the author, not the institution.

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index